基于概念和段落檢索的生物醫(yī)學文獻知識發(fā)現(xiàn).pdf

1、大連理工大學碩士學位論文基于概念和段落檢索的生物醫(yī)學文獻知識發(fā)現(xiàn)姓名：陳冉申請學位級別：碩士專業(yè)：計算機系統(tǒng)結構指導教師：林鴻飛20081201大連理工大學碩士學位論文ConceptandPassageRetrievalforKnowledgeDiscoveryfromBiomedicalLiteratureAbstractExtractingtheimplicitbiologicalrelationshipfrombiomedical

2、literaturecontributestothebiomedicalhypothesiswhich伽beexploredthroughtheexperimentsfurtherHowever’withthegrowthofthenumberofbiomedicalliterature。itisimpossibleteadalltheliteratureartificiallyDonRSwanson，thepmfessoremerit

3、usinthedepartmentofinformationscienceattheUniversityofChicagepresentednon—interrelatedliteraturebasedknowledgediscoverymethodwhichcanexplorethehiddenconnectionfromlargebiomedicalliteratureandformthehypothesistoleadthesci

4、enceexperimentSwanson’SknowledgediscoverymethodcomestobeaveryhotresearchtopicnecalculatingprocessofconceptretrievalissimpleandintuitionisticInthepaperCOoccurrenceanalysistechnologyisusedtoextractpotentialknowledgefromthe

5、biomedicalliteraturedatabase——M匣DL]小i]BThcexperimentfocusesondiscoveringnewbiologicalconnectionsbetweendiseaseandchemicalsdrugsorgenes11lreecomputationalmethodsarecomparedinthepaperforscoringandrankingtheMesHterms：zscore

6、，徹F(TermFrequencyInverseDocumentFrequency)andPMI(PointwiseMutualInformation)wcreportonthreesetsofexperiments：Alzheimer’Sdisease，Migrainedisorders，SchizophreniaandusetheinformationretrievalmetricswhichisintroducedbyLitLin

7、kersystemtoevaluatctheperformancesAccordingtothecharacteristicsofthethreemethods，afusionformulaisintroducedforrerankingandrechoosingthetermstoimprovethefinaloutcomeWefoundthatzsoDreand田FIDFshowbetterperformanceindifferen

8、texperimentsandourempiricalresultsvalidatetheeffectivenessoffusionapproachThediscoveredresultscontainmoretendencytopicwhichmeetstheknowledgediscoveryrequirementPassageretrievalisintroducedforthefurtherexperimentbecauseth

9、eresultdataofconceptretrievalknowledgediscoveryaredispersiveandithasnegativeimpactontheperformanceofthefinalrankingInthenextexperiment，passageretrievalisusedtoextracttheMeSHconceptinthecompletesentencesfromtheabstractofM

10、EDLINETocomparetheresultswiththeresultsofconceptretrievalexperimentthesamecomputationalmethodandstartingtopicareusedintheexperiment，andevaluatetheprecisionandrecallTheresultsshowthattheMeSHconceptswhicharediscoveredbythe